Enhancing Soil Moisture Forecasting Accuracy with REDF-LSTM: Integrating Residual En-Decoding and Feature Attention Mechanisms

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This paper proposes a novel deep learning model called REDF-LSTM for predicting soil moisture and evaluates its performance on benchmark datasets. The topic of improving soil moisture prediction is highly relevant, and the proposed methodology integrating residual learning, encoder-decoder architectures, and attention mechanisms is interesting. However, there are some issues that need to be addressed:

1-     The model description and motivation for different components need more clarity. The encoder-decoder and attention mechanisms are not clearly explained in the context of soil moisture prediction. More details on why these are suitable for time series modeling are required.

2-     The experiment section lacks important details on hyperparameters, training procedures, ablation studies etc. Without these, it is difficult to properly evaluate the model and results.

3-     The comparison to other models is limited. More advanced baselines from the literature should be used instead of vanilla LSTM models. Details on experimental settings for different models are missing.

4-     The results reported lack quantitative comparisons. Performance metrics on test datasets with error bars need to be provided to substantiate the claims of improved accuracy.

 

5-     Discussion on limitations and scope for future work is brief. More analysis is required to link the findings back to the problem and identify open challenges.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Enhancing Soil Moisture Forecasting Accuracy with REDF-LSTM: Integrating Residual En-Decoding and Feature Attention Mechanisms

 

Comments to Authors

Update the introduction section of the article with details regarding the specific challenges in soil moisture prediction. Do traditional models face REDF-LSTM addresses more effectively?

In section 2 of the article can you yell how does the integration of the residual learning encoder-decoder and feedforward attention mechanisms in REDF-LSTM enhance its predictive accuracy compared to traditional LSTM models?

Improve labels (y-labels of bars) of Figure 1, they are very difficult to read.

Can you elaborate on the data preprocessing steps and the type of input data used to train the REDF-LSTM model mentioned in Eq 1 to 6.

In section 2.2.3, please update the limitations to the REDF-LSTM model, and if so, how might future research address these?

How does the feedforward attention mechanism specifically contribute to identifying key influencing factors in soil moisture prediction and what are the potential applications of the REDF-LSTM model in precision agriculture and ecosystem management.

In section 2.2.5 you can see some bold italic letters remove the bold, increased the font size of captions in Figure 5. Same in Figure 6, 7, 8  increase the font size.

In the section 3.3 update, how does the REDF-LSTM model handle varying climatic conditions across different geographic regions in its predictions? Could the methodologies and innovations introduced in the REDF-LSTM model be applied to other fields of environmental prediction or data analysis?

 

Discussion and conclusion seem convincing.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors addressed my comments, so I recommend accepting the paper.